A math problem about proability fusion in sensor fusion module

[LC041336]

What is the math method or principle for "welsh_var_loss"??? If somebody knows it, please give me a reference link. Thanks a lot!!!
'''
double WelshVarLossFun(double dist, double th, double scale)
{
double p = 1e-6;
if (dist < th)
{
p = 1 - 1e-6;
}
else
{
dist -= th;
dist /= scale;
p = std::exp(-dist * dist);
}
return p;
}

'''